Page 12 of 39 FirstFirst ... 2101112131422 ... LastLast
Results 111 to 120 of 382

Thread: Nvidia's Fermi GTX480 is broken and unfixable

  1. #111
    Quote Originally Posted by p4man View Post
    idle/loaded has nothing to do with this, we're talking TDP, so loaded.
    And Ive never seen a gpu or cpu operate at 0 Mhz either
    really? i know my 5850 is running in a power state that downclocks to 0Mhz at this very second

  2. #112
    8-bit overflow
    Join Date
    Jul 2009
    Posts
    426
    Quote Originally Posted by AirfarceNone View Post
    really? i know my 5850 is running in a power state that downclocks to 0Mhz at this very second
    What can I say.. thats amazing Unless of course, its not operating

  3. #113
    Quote Originally Posted by p4man View Post
    Maybe Im missing your point, but power draw scales linear with clockspeed (if you keep voltage constant) and to the square of the voltage (and voltage can be scaled down with clockspeed, although not linearly obviously).
    That hasn't been true for quite some time. Static power is a significant contributor to power in most designs of any appreciable size. In a design the size of G100 its likely that dynamic power is less than half of the total power.
    speaking for myself inc.

  4. #114
    8-bit overflow
    Join Date
    Aug 2009
    Posts
    377
    It has everything to do with it, as just like wide open, idle also needs a minimum for stability/take up.
    You need to keep it active, whther its full usage or idle, its like adding enough gas into your chamber, not enough, it wont fire, too much, itll burn

  5. #115
    Quote Originally Posted by p4man View Post
    What can I say.. thats amazing Unless of course, its not operating
    sure its operating, chugging along nicely at 0

  6. #116
    >intel 4004
    Join Date
    Jun 2009
    Posts
    11,281
    Very Low Power State is not at 0 MHz for the GPU, should be your GPU information software (e.g. GPU-Z) report a wrong number of core frequency.

  7. #117
    8-bit overflow
    Join Date
    Aug 2009
    Posts
    377
    I like Intels and AMDs new ability to route around an entire core, to effectively shut it down, to eliminate this, and create a virtual no power idle state, saving it all in cache for take up, which is always on the active core. No leakage, no minimals

  8. #118
    >intel 4004
    Join Date
    Jun 2009
    Posts
    11,281
    Now for everybody who suspects the article disappeared from the main page, just take a closer look under the three boxes under the date 2010-02-17 (YYYY-MM-DD):

    Wed 17 Feb 2010
    NVIDIA'S FERMI GTX480 IS BROKEN AND UNFIXABLE

    Hot, slow, late and unmanufacturable
    It's only a matter of site layout, when new articles published, and the number of articles for a particular day is more than 3, then the old articles will be pushed out of the three boxes, which is used to highlight the latest articles of that day.

    So, no worries, since Charlie won't retract any articles, unlike Theo...

  9. #119
    Quote Originally Posted by aaronspink View Post
    That hasn't been true for quite some time. Static power is a significant contributor to power in most designs of any appreciable size. In a design the size of G100 its likely that dynamic power is less than half of the total power.
    NV process engineers must have known that they were walking near or on the edge with G100 in relation to leakage and power. Knowing the huge R&D budget for NV, i simply dont understand there is no minor dirivatives of this suposedly modular GPU (Rys statement). Why? - And where is the backup product?
    I mean, this is not about investment in GPU market only, they have an huge investment in
    A. software portfolio
    B. Brand
    And then no alternative strategies to protect these investment? Its risk management and performance mangement so bad, it just does not make sense. What do I not see?

  10. #120
    8-bit overflow
    Join Date
    Jul 2009
    Posts
    426
    Quote Originally Posted by aaronspink View Post
    That hasn't been true for quite some time. Static power is a significant contributor to power in most designs of any appreciable size. In a design the size of G100 its likely that dynamic power is less than half of the total power.
    Really? Half?

    Hmm. I just looked up some numbers for cypress, comparing TDP with iddle power at 1/5th the speed, and you might actually be right. I had no idea it was that bad.

    Here is what I did:
    cypress: 188W TDP. If half of that were static power, that would mean at least 94W iddle at full clock speed, assuming somehow dynamic power would be zero while iddling. At iddle cypress runs the core at 150 Mhz instead of 850, so you'd expect at the very least 16W (ignoring a lot of stuff, including RAM, but that is clocked down by comparable amounts).

    in reality it seems to use 20W when doing absolutely nothing (second x-fire card idling, Im assuming its still refreshing its ram and running its cores at 150 MHz). Or maybe one should use the 27W iddle power instead, but either number suggests you're right. Didnt know it was that bad. SOI to the rescue ?

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  
WordPress Appliance - Powered by TurnKey Linux